arctic ocean sea ice snow depth evaluation and bias ... · abstract. sea ice cover in the arctic...

The Cryosphere, 7, 1887–1900, 2013www.the-cryosphere.net/7/1887/2013/doi:10.5194/tc-7-1887-2013© Author(s) 2013. CC Attribution 3.0 License.

The Cryosphere

Open A

ccess

Arctic Ocean sea ice snow depth evaluationand bias sensitivity in CCSM

B. A. Blazey1,2, M. M. Holland 3, and E. C. Hunke4

1Center for Marine Environmental Sciences (MARUM), Universität Bremen, Bremen, Germany2Department of Atmospheric and Oceanic Sciences, University of Colorado, Boulder, Colorado, USA3National Center for Atmospheric Research, Boulder, Colorado, USA4Los Alamos National Laboratory, Los Alamos, New Mexico

Correspondence to:B. A. Blazey ([email protected])

Received: 1 March 2013 – Published in The Cryosphere Discuss.: 9 April 2013Revised: 21 October 2013 – Accepted: 26 October 2013 – Published: 13 December 2013

Abstract. Sea ice cover in the Arctic Ocean is a continuedfocus of attention. This study investigates the impact of thesnow overlying the sea ice in the Arctic Ocean. The impact ofsnow depth biases in the Community Climate System Model(CCSM) is shown to impact not only the sea ice, but also theoverall Arctic climate. Following the identification of sea-sonal biases produced in CCSM simulations, the thermody-namic transfer through the snow–ice column is perturbed todetermine model sensitivity to these biases. This study con-cludes that perturbations on the order of the observed biasesresult in modification of the annual mean conductive fluxthrough the snow–ice column of 0.5 W m2 relative to an un-modified simulation. The results suggest that the ice has acomplex response to snow characteristics, with ice of differ-ent thicknesses producing distinct reactions. Our results in-dicate the importance of an accurate simulation of snow onthe Arctic sea ice. Consequently, future work investigatingthe impact of current precipitation biases and missing snowprocesses, such as blowing snow, densification, and seasonalchanges, is warranted.

1 Introduction

The decline of Arctic Sea ice extent over recent decades iswell documented (Parkinson et al., 1999; Meier et al., 2007).This decline continues unabated, with the period 2007–2013setting records for six of the seven lowest summer ice extentssince the beginning of the satellite record (NSIDC, 2013).Last summer, 2012, set the record for the lowest summer ice

area (NSIDC, 2012). The high sensitivity to climate changein the Arctic (Holland et al., 2006a, b) combined with an-ticipated feedback mechanisms, such as the ice albedo feed-back (Curry et al., 1995), warrants a continued focus on real-istically simulating the Arctic system in large-scale climatemodels. In addition, while the model representation of thetransition from a perennial to a seasonal ice pack is improv-ing, this shift continues to progress more quickly than sim-ulated by many general circulation models (Stroeve et al.,2007, 2012). Major shifts in the character and age of the iceare indicative of a persistent shift in the state of the Arctic icecover (Maslanik et al., 2007, 2011). These changes include ashift to a younger, thinner, and more seasonally variable icepack. As the ice in the Arctic Ocean becomes more simi-lar to the ice in the Southern Ocean, it may be that featuressuch as snow depth will increase in importance to the Arc-tic sea ice (Fichefet and Maqueda, 1999), and in turn to theregional climate. In addition to changes in the ice itself, theoverlying snowpack is projected to decline in the Arctic inseveral GCMs (Hezel et al., 2012; Vavrus et al., 2012). Thesechanges are projected to significantly impact the thermody-namics of the sea ice (Blazey, 2012). This highlights the needto examine thermodynamic processes controlling the evolu-tion and state of the Arctic ice pack.

The thermal conductivity of snow is nearly an order ofmagnitude less than the thermal conductivity of the ice itcovers, and as such snow depth is an important componentof the thermal transfer through the ice column despite its rel-atively tenuous nature. In previous sensitivity studies usingmodels, the net effect of snow depth on the ice pack varies.

Published by Copernicus Publications on behalf of the European Geosciences Union.

1888 B. A. Blazey et al.: Arctic Ocean sea ice snow depth evaluation

One of the earlier model sensitivity studies of sea ice foundthat snow on sea ice has two primary competing effects onthe ice mass budget (Maykut and Untersteiner, 1971). Dur-ing the ice growth season, the presence of snow insulates thesea ice, reducing the transfer of heat from the ice–ocean inter-face. This in turn reduces ice growth. However, in the tran-sition to the surface ice melt season, the presence of snowcan delay surface ice melt and lead to a reduction in the sea-sonal ice mass loss. The high albedo of snow relative to seaice also plays an important role in the surface heat budgetsand the onset of ice melt. While Maykut and Untersteiner(1971) identified the major processes by which snow modu-lates the ice thermodynamics and mass balance, they foundthat in the idealized model these effects were balanced forsnow depths ranging from 0 to 70 cm. However, this findingdid not hold with subsequent studies. Later studies, under-taken both in situ and using various modeling environments,have found a variety of responses to snow depth.

For example, Brown and Cote (1992) found ice thermo-dynamics, including ice growth, to be highly sensitive tosnow cover, while Holland et al. (1993) found snow to beof secondary importance to ice characteristics. Fichefet andMaqueda (1999) found the ice cover in the Southern Ocean tobe “remarkably sensitive to the accumulation rate of snow”.Massom et al. (2001) observed the highly complex and vari-able role of snow on Antarctic sea ice. The authors concludedthat a proper treatment of snow on sea ice is important to cli-mate modeling. Cheng et al. (2008) found snow depth to bean important element in the sea ice system within a modelingenvironment. While more recent studies have explored theintroduction of additional thermodynamic complexity to thetreatment of snow (Lecomte et al., 2011; Cheng et al., 2008),these studies have focused on one-dimensional modeling andevaluation.

The role of snow cover in modulating ice growth has beenverified in the field, where the heterogeneity of snow dis-tribution has been observed to produce corresponding het-erogeneous growth in underlying lake ice (Sturm and Lis-ton, 2003). More specifically, focused snow thermal con-ductivity sensitivity studies have found a 10–20 % changein ice thickness when snow thermal conductivity is halvedfrom the typical∼ 0.3 W m−1 K−1 to the measured valueof 0.15 W m−1 K−1 (Sturm et al., 2002; Wu et al., 1999;Fichefet et al., 2000).

Considering the range of findings in previous investiga-tions of the net role of snow depth to the sea ice mass ther-modynamics, we are motivated to investigate this aspect ofthe Arctic climate system in one of the major current GCMs.As such, this study evaluates the snow depth and density oversea ice as simulated in Community Climate System Model(CCSM) fully coupled runs using in situ measurements ofsnow depth. Sensitivity runs are then performed to determinethe influence of snow depth and density biases on both the icestate and the Arctic climate. Additional evaluations of snowcharacteristics and simulations of the stand-alone ice model

the Community Sea Ice Code (CICE) were also performedand are reported in Blazey (2012). These simulations foundconsistent ice sensitivity to decreases in snow depth and amean state-dependent response to increases in snow depth.

Due to the coupled nature of these simulations, we expectinitial perturbations to the ice caused by changes to the snowcharacteristics to result in feedbacks with other componentsof the Arctic climate system.

Section 2 describes the GCM utilized: CCSM4 and itscomponent ice model, CICE. Section 3 describes the in situmeasurements and evaluation method. Section 4 discussesevaluation of CCSM, including possible causes of snowdepth biases identified therein. Section 5 details the designand results of the CCSM bias sensitivity experiment. Sec-tion 6 begins with a discussion of the implications of ourresults with regards to the CICE and the CCSM results as awhole. We conclude with a discussion of future directions forthis work.

2 Model: CCSM4

CCSM4, released in Spring 2010, includes coupled compo-nent models for the atmosphere, ocean, ice, and land surfaceand is mass and energy conserving. This study makes use ofCCSM in two configurations. For the evaluation portion ofthis study, the release version of the CCSM4 is used with allcomponents active. The sensitivity portion of this study uti-lizes a reduced version of CCSM with the fully active oceanreplaced by a slab ocean model (SOM). Here we focus on thesea ice component model and improvements to CCSM mostimportant to the Arctic Ocean. A more comprehensive dis-cussion of the improvements in CCSM4 is available in Gentet al. (2011).

The sea ice component model used in CCSM4 is the LosAlamos Sea Ice Model (CICE) version 4 (Hunke and Lip-scomb, 2010). CICE is designed to be computationally effi-cient while including a thermodynamic model, a model of icedynamics, a transport model that describes advection, ridgingparameterization, and a sub-grid-scale ice thickness distribu-tion (Thorndike et al., 1975). CICE includes elastic–viscous–plastic dynamics (Hunke and Dukowiz, 2002) and is energyconserving (Bitz and Lipscomb, 1999). In the default con-figuration, CICE has a 30 min time step, 4 ice layers and 1snow layer; however these values can be changed. In the con-figurations used in this study, the sub-grid-cell ice thicknessincludes five prescribed thickness categories. Each categoryincludes a discrete thermodynamic treatment including snowcalculations. These calculations are uniform across each gridcell at each time step. Ice and snow are transferred betweencategories by means of a one-dimensional linear remapping(Lipscomb, 2001) and include deformation, dynamical ad-vection, and thermodynamic thickness change as pathways.Spatial advection occurs by means of a two-dimensional lin-ear remapping (Lipscomb and Hunke, 2004).

The Cryosphere, 7, 1887–1900, 2013 www.the-cryosphere.net/7/1887/2013/

B. A. Blazey et al.: Arctic Ocean sea ice snow depth evaluation 1889

In general, the thermodynamic and dynamic treatment ofsnow follows the treatment of the ice. However, snow is onlytransferred between ice thickness categories or grid cells withthe ice it overlies. As such, the snow overlying the sea ice canbe considered a feature of the ice, rather than an independentprocess. However, there are a handful of processes where thesnow is treated separately. Other than precipitation and meltthese include snow-to-ice formation and loss of snow massduring ice deformation.

Recent improvements to CICE include a delta-Eddington(DE) multiple scattering treatment of snow and ice, whichuses internal snow properties to determine albedo and setsCICE ahead in relation to most GCMs (Briegleb and Light,2007). In the DE parameterization of CICE, surface albedois computed using the thickness of snow and ice, in combi-nation with a melt pond layer. Parameters such as snow grainsize are also included, and the treatment compares favorabil-ity with in situ measurements. However, neither temperaturenor snow age is considered. This may potentially lead to bi-ases in the surface albedo, considering the large changes insnow albedo as snow ages and melts (Perovich et al., 2002,2007). The melt and aging of snow and sea ice lead to largealbedo decreases, upward of a factor of 2 in the case of snowalone (Perovich et al., 1998). However, investigating the im-pact of snow and albedo treatment biases is beyond the scopeof this work. For purposes of this study we consider the delta-Eddington treatment to be an adequate albedo. CICE nowincludes a non-zero heat capacity for the snow cover, meltponds, and aerosol deposition. The effects of this parameter-ization are reported in Holland et al. (2012).

Despite these improvements, the density and thermal con-ductivity of snow in CICE are fixed. As a result, the treat-ment of CICE snow is more rudimentary than the treatmentof snow in the Community Land Model (CLM) (Lawrenceet al., 2011; Olsen et al., 2010). Additionally, CICE does notcurrently include blowing snow, which has been found to sig-nificantly reduce snow depth biases in a focused snow modelenvironment via snow loss to leads (Chung et al., 2011; Déryand Tremblay, 2004). However, the advection of ice coverpresents a distinct challenge for snow treatment by CICE incomparison to CLM, and is not included in most model en-vironments. Moreover, the actual snow depths produced insimulations including CICE have not been validated.

The atmospheric component model, the Community At-mospheric Model version 4 (CAM4), includes 26 verticallayers with a resolution of 1.25◦ by 0.9◦. CAM4 uses a Lin-Rood dynamical core (Lin, 2004). For a more detailed de-scription of CAM4, see Neale et al. (2013). In the Arctic, theinclusion of a freeze dry modification serves to reduce thewinter low-level clouds in the Arctic (Vavrus and Waliser,2008).

The land model, the Community Land Model version 4(CLM4), uses the same grid as CAM4, and includes improve-ments to hydrology. CLM4 makes use of the Snow and IceAerosol Radiation model (SNICAR) (Flanner et al., 2007),

and includes grain-size-dependent snow aging, aerosol de-position, and vertically resolved snowpack heating. For fulldocumentation, see Lawrence et al. (2011) and Oleson etal. (2010).

The Parallel Ocean Program version 2 (POP2) is the oceancomponent of CCSM4. POP2 uses a 1-degree grid with theNorth Pole displaced into Greenland. POP2 uses 60 verticallevels, with a surface layer thickness of 10 m, increasing withdepth (Smith et al., 2010). Improvements to POP2 includemixing parameterizations allowing improved simulation ofNorth Atlantic dense water transport (Briegleb et al., 2010;Danabasoglu et al., 2010).

In order to save computation time, the fully active POP2is replaced with a SOM in the sensitivity experiments de-scribed in Sects. 4 and 5. The SOM serves as a heat reservoir,while calculating a surface ocean temperature for a fixed-depth layer using a surface energy budget calculation and aprescribed ocean heat transport (Bitz et al., 2011). This pre-scribed transport is obtained from coupled simulations (Bai-ley et al., 2012). As a result, additional processes such ascurrents and oceanic transport are omitted. Our simulationsuse the SOM rather than the full POP2 in large part due tothe lower computational costs, a result of both the simpli-fied model and the faster equilibration of the model. Danab-soglu and Gent (2009) found that the results of the SOM andfully coupled simulations were similar under control condi-tions, with differences under a perturbed climate focused inthe Southern Ocean.

The model output hereafter referred to as CCSM makesuse of a six-member ensemble of fully coupled 20th-centurysimulations (Gent et al., 2011). Each ensemble member issimulated over the period 1850–2005 and includes anthro-pogenic effects for this period. The general performance ofCCSM4 in this ensemble is available in Gent et al. (2011).More specifically, discussions of the Arctic region perfor-mance for this ensemble are available in the CCSM specialcollection of theJournal of Climatefor atmosphere (de Boeret al., 2012) and sea ice/ocean (Jahn et al., 2012) conditions.

3 Snow depth evaluation

3.1 In situ data

The primary source of in situ snow measurements is Rus-sian drift stations (Arctic Climatology Project, 2000). Thesestations were located on multiyear ice, and at least one sta-tion was in operation during the period 1954–1991. Snowdepth measurements were made in two ways. First, an ide-ally daily observation occurred at a snow stake adjacent tothe ice camp. Second, once to thrice monthly, a 500 m to1000 m transect “snow line” was made at least 500 m fromthe camp. Snow depth was measured every 10 m along thistransect. In addition to depth, snow density was measured us-ing a cylinder massing technique. Transects were made when

www.the-cryosphere.net/7/1887/2013/ The Cryosphere, 7, 1887–1900, 2013


snow depth exceeded 0.05 m and covered at least 50 % ofthe transect length. Subsequent transects would occur in thesame direction but offset several meters. When making useof transect snow measurements, we consider the mean of atransect value (depth or density) to constitute a single mea-surement. While it would be natural to assume the snow stakemeasurement could be contaminated by snow alteration dueto proximity to the ice camp, Warren et al. (1999) did notfind this to be the case. However, Warren et al. (1999) didfavor using the transect snow depths when generating a snowclimatology. While the in situ measurements are used for thepoint-by-point model data comparison, this study does makeuse of the Warren et al. (1999) snow climatology when com-paring the geographic distribution of snow depth.

Ice mass balance (IMB) buoys managed by the Cold Re-gions Research and Engineering Laboratory (CRREL) wereconsidered as an additional source of in situ snow depth data(Perovich et al., 2009). However, the buoys’ advection pathsalong the eastern coast of Greenland were found to cause dif-ficulty matching the buoys to appropriate model grid cells.For a discussion of this issue see Blazey (2012).

3.2 Evaluation method

Due to using an ensemble of fully coupled historical forc-ing simulations (hindcasts), we do not expect that the modelwill capture the depth of snow in a given year in the in situmeasurements. Instead, we examine the ability of the modelto produce both reasonable mean monthly snow depths, andreasonable variation from the mean in comparison to the insitu measurements. As such, we regard the hindcast snowdepths and in situ measurements to be independent samples,with each sample being a snow depth in a given month. Inthe method that follows, we use a selection process to com-pare ice of similar geographical location, historical period,and characteristics to the in situ measurements.

In the model ensemble used, monthly mean model statesare saved. As such, in situ measurements discussed in theprevious section are matched to a single monthly output file.In this way, the sample size of the snow line measurementsand matched model grid cells are equal. In the case of thesnow stake measurements, the drift of the station results in1 to 30 snow stake measurements per matching grid cell. Inboth cases, we compare the monthly mean snow depth forthe period 1954–1993.

A given in situ measurement is compared to the model gridcell with the closest center point, with the following selectioncriteria. The matched CCSM locations are restricted to gridcells above 70◦ N, which eliminates the seas along the mar-gin of the Arctic. Each in situ measurement is matched to twomodel grid cells. The first is the nearest grid cell, polewardof 70 N, with at least 15 % ice cover, hereafter referred to asthe “all_ice” snow thickness, as all ice cover is considered.This definition of ice cover corresponds to commonly usedice extent definitions (ex: Meier et al., 2007). However, 15 %

ice cover does not correspond well with the ice conditions inwhich the in situ measurements were made since the stationswere located in areas of high-concentration, multiyear ice.To better match the thicker multiyear ice conditions of theRussian stations, this matching process is repeated with theselection restricted to the nearest grid cell, north of 70 N, withat least 50 % of the cell covered with ice greater than 1.5 mthickness. We refer to this snow depth as the “thick_ice”depth. A thickness of 1.5 m is used because it is the max-imum thickness of the second of five thickness categories.In general, the exact choice of thickness does not affect thesnow depth.

In the case of monthly snow line measurements, the meanof the transect is taken as the sample, resulting in equal sam-ple sizes for the in situ measurement and model values. Inthe case of the daily snow stake measurements, the drift ofthe station itself results in multiple model grid cells corre-sponding to a given set of monthly measurements. In a repre-sentative hindcast, if all_ice matching is used there is a meanof 4.5 in situ measurements per matched model grid loca-tion. If the thick_ice matching is used, there are 4.4 in situmeasurements per model location. In general, restricting thematching to model locations with thick_ice results in slightlygreater distances between the matched in situ and model lo-cations. The distance increases a mean of 17 km to 23 km inthe case of the monthly transects, and from 21 km to 25 kmin the case of the daily stake measurements.

Using these two methods, a reasonable comparison ofCCSM-produced monthly mean snow depths for the period1954–1993 is performed. Following the matching process wehave matched data sets of snow depth: transect to all_icethicknesses, transect to thick_ice, and similarly for snowstake measurements. A paired Student’st test was performedon each pair by month. Again, we do not expect the samplesto be exactly matched due to the model’s inability to simulatea given year. However, by matching time of year, geographi-cal location, and year (mostly for era, not particular year), weproduced paired samples. The degrees of freedom vary sig-nificantly, from as little as 3 in the summer for the transect to2000 for the snow stake measurements. With the exceptionof summer, the DOF for the transect is approximately 50 foreach month, and 2000 for the snow stake.

3.3 CCSM snow evaluation results

Unsurprisingly considering the excessive precipitation pro-duced by CAM in the Arctic (de Boer et al., 2012), and lackof blowing snow in CICE, the comparison indicates signifi-cant snow depth and density differences between the CCSMensemble and the in situ measurements. Figure 1 shows themodel snow depths in relation to the in situ measurements.At a qualitative level, CCSM4 snow depths are generally toothick relative to the in situ measurements.

CCSM snow depths are 20 % in excess of Russian driftstation transects on all_ice, and 30 % in excess on thick_ice.



These excesses are much higher by midsummer, and in Au-gust reach 80 % and 105 %, respectively. The average meanannual biases in absolute terms are 4.7 cm and 6.4 cm, whilethe August values are actually higher, 6.9 cm and 7.7 cm, forall_ice and thick_ice, respectively. The differences betweenthe CCSM snow depths (for both all_ice and thick_ice) ofthe individual ensemble members in Fig. 1 are not necessar-ily statistically significant (p <= 0.05) in comparison to theRussian in situ data. When considering individual ensemblemembers, the most likely months to lack statistically signif-icant differences in snow depths are June, July, and August.This is likely due to a low number of in situ samples for thesummer period. However, taken as an ensemble, the differ-ences for the CCSM snow depth differences (both all_ice andthick_ice) are statistically significant for all months exceptJuly when validated against the Russian drift station transectdata.

In addition, the Russian drift station transect data has alower standard deviation in snow depth than the model, amean of 30 % of the total snow thickness, in comparison to54 % and 49 % for the corresponding all_ice and thick_iceCCSM snow thickness. It is unsurprising that the all_icecomparison produces the largest variance, as ice of highlyvariable thickness is selected. As a result, thin ice that hasaccumulated little snow is included in this population. Thesnow stake average standard deviation, 64 %, is larger in partdue to each value being a single measurement, rather thana mean of many measurements. The two in situ variancesbracket the 51 % and 48 % mean standard deviations for thecorresponding CCSM all_ice and thick_ice snow depths, asdetailed below. Because the CCSM snow depths are aver-aged over a grid cell, rather than the point measurements ofthe snow stake in situ measurements, the lower standard de-viations for these snow depths are unsurprising.

CCSM snow depth biases are slightly higher in relationto the Russian drift station snow stake measurements. Theall_ice snow depth excess is greater than 20 %, and thethick_ice snow depth excess is greater than 50 %. As withthe snow line, the modeled snow biases relative to the snowstake measurements are very high in the summer, exceeding80 % and 195 % in August. The significance of these compar-isons is very similar to the transect comparisons: all monthsexcept July satistfy our significance test (P < 0.05)

Next, the geographical distribution of the thickness biasis examined, as shown in Fig. 2. In general, it appears thatthe CCSM ensemble produces a snow depth excess of about20 cm near the Canadian Arctic Archipelago when comparedto the Warren et al. (1999) climatology. This excess is mostpronounced in the winter and summer (as opposed to the fallfreeze-up) when CCSM produces snow greater than 50 cmthick and the in situ snow climatology is∼ 30 cm thick inthis region. However, note that this occurs in the regions ofthickest ice, along the Canadian Arctic Archipelago. There isa lack of Russian drift stations used by Warren et al. (1999) inthis region, as shown by the symbols in Fig. 2. However, ex-

trapolation of the snow depths from central Arctic stationsindicates thicker snow does accumulate in the vicinity ofGreenland and the Canadian Arctic Archipelago. The snowclimatology reproduced in the bottom row of Fig. 2 suggeststhe thickest snow was observed on this side of the Arctic. Itseems possible that very deep snow does occur in this regionof the Arctic Ocean, but additional station coverage wouldbe required to detect this feature. Regardless, while CCSMis producing too much snow depth overall, the highest biasesare in the area where the thickest snow depths are expected.

Finally, Fig. 3 documents density bias between the mea-surements made as part of the Russian snow depth transectand the CICE default snow density of 330 kg m−3 (Hunkeand Lipscomb, 2010). Note that no snow density measure-ments were made in July and August. The most apparent dif-ference results from the omission of snow densification bythe invariant snow density in CICE. As a result, the snowdensity is approximately 30 % too great in the early autumn,but is approximately correct by the onset of melt. This im-plies that the excess in snow mass, or snow water equivalent,is larger in the fall than the depth comparison would imply.

Overall, we find that the CCSM ensemble produces year-round excesses in snow depth, for both of the in situ mea-surements (stake and transect) used for evaluation of CCSM.Two potential causes of this excessive snow depth come tomind. First, the CAM produces excessive Arctic precipitation(de Boer et al., 2012). Second, CICE does not treat the lossof blowing snow to leads and sublimation, which was foundto reduce high biases in modeled snow depth (Chung et al.,2010). The excess is especially pronounced in the summermonths. In addition to depth biases, the use of a fixed snowdepth parameter results in snow that is too dense in the fall,but is much nearer the in situ measurements by the onset ofthe spring and summer melt.

4 Bias sensitivity simulation

4.1 CCSM bias sensitivity simulation design

The sensitivity portion of this study seeks to determine theeffects of snow depth and density biases on ice characteris-tics and the controlling thermodynamics. Biases in the snowcover will impact the insulating properties of the sea ice,with implications for ice growth and melt. Both are impor-tant processes that can lead to biases in the simulated seaice mass budgets. The general method is intended to assessthe effects of snow depth changes of a magnitude equal tothe bias we reported in Sect. 3.3 on the Arctic ice cover andclimate. Specifically, we modify the thermal conductivity ofthe snow in CCSM/CICE to account for the observed biasin snow depth. As such, the ice will experience the snow ashaving the thermal conductivity of a thinner snowpack. Snowconductivity has previously been shown to have an impact onsea ice growth in a model environment (Fichefet et al., 2000).



Fig. 1.Panels(a) and(b): CCSM Arctic Ocean snow depths for the period 1954–1993 validated against Russian drift station measurements.Panel(a) displays the Russian snow line as the in situ data; panel(b) uses the Russian snow stake as the in situ data. Each ensemble memberis reported separately. Black squares indicate the in situ values for the period; red diamonds indicate the snow depth overlying the nearest iceof any thickness; blue asterisks indicate the snow depth overlying the nearest ice with greater than 1.49 m thickness. Vertical bars indicateone standard deviation for both in situ and model snow depths.

Fig. 2.Geographical distribution of modeled snow depth and in situmeasurements, in cm. As defined by the color bar, red correspondsto low snow depth and blue to high snow depth. From left to right:winter (January, February, March), summer (June, July, August),and freeze-up (October). The top row is the CCSM ensemble snowdepth climatology. The second row overlies the position of the Rus-sian drift station locations over a snow depth climatology derivedfrom the in situ measurements taken at these stations (Warren et al.,1999).

While the basic principle has previously been documented,this study extends previous work by investigating the role offeedback processes between the sea ice, atmosphere, and sur-face ocean.

While altering density and snow thickness directly wouldbe the most straightforward method to explore biases inthese quantities, such alterations pose difficulties surround-ing mass and energy conservation. More specifically, reduc-ing the precipitation mass at the coupling interface between

Fig. 3. Snow density bias: CICE model default less in situ mea-surements; bars indicate standard deviations (in the in situ measure-ments.

CAM and CICE would result in the removal of precipita-tion mass and the thermal energy present in the mass fromthe model environment. Instead, to assess the implicationsof the snow biases for sea ice thermodynamics, we manip-ulate the snow thermal conductivity. This method allows usto investigate the thermodynamic role of snow in the thermaltransmittance through the snow–ice column. To better under-stand this process, it is useful to consider thermal transmit-tance,Utotal, in Eq. (1), which is the rate at which energyis transferred through a barrier of a given thickness understeady-state conditions. However, note that this is not a partof the CICE model, and is used here only for clarity. Thermaltransmittance is useful to consider here because there are twobarriers present in the sea ice system, ice and snow. Thermalconductivity,k, is the rate of transfer through a barrier per



unit thickness,h, and is a characteristic of the material. InEq. (1) h/k can be considered a thermal resistance, and iscumulative as in Eq. (1). In Eq. (1),Utotal is the total ther-mal transmittance, here a function of the snow and ice ther-mal conductivities (ksnow,kice) and thickness (hsnow,hice). Itis useful to consider Eq. (1), the control situation.

Utotal =1

hicekice

+hsnowksnow

(1)

For purposes of this sensitivity study, we wish to adjust thevalue ofksnow to ksensitivity, a conductivity used to make thesnow appear “correct” thermally, as in Eq. (2).

Usensitivity=1

hicekice

+hsnow

ksensitivity

(2)

We adjust the conductivity such that the snow thermal trans-mittance accounts for the bias in model snow depth and den-sity described earlier. Beginning with depth adjustments, wetake the product ofksnow andhinsitu/hsnow(eval), wherehinsituis taken from the in situ measurements, andhsnow(eval) is theCCSM snow depth matched to the in situ measurements. Theadjusted snow thermal conductivity reported in Eq. (2). Byway of example, if the CCSM snow is too thick, the snowthermal conductivity is effectively reduced to adjust for this,causing the thermodynamics to experience a thinner snowcover.

kh sensitivity= ksnowhsnow(eval)

hinsitu(3)

Next, Eq. (2) is adjusted by the addition of aρinsitu/ρsnowdensity values for the in situ measurements and default CICEsnow densities, effectively adjusting the effective snow depthto account for density biases. The final modified conductiv-ity is reported in Eq. (4). This conductivity is applied in theCCSM conductivity experiment.

ksensitivity= ksnowhsnow(eval)

hinsitu

ρinsitu

ρsnow(4)

An average of the monthly bias derived from the snow stakebias and snow line bias is used to calculate the biases testedin this section. This is a simpler method than that used byWarren et al. (1999) in the development of their climatology,but will serve to determine whether the biases are relevant tothe state of the Arctic ice.

In addition, rather than consider the direct effect of snowdensity on conductivity as determined by Sturm et al. (1997),the bias correction instead compensates for the excessivesnow density early in the season by mimicking a thickersnowpack, which would contain the same snow mass per unitarea while having a lower density. We consider a linear rela-tionship between excessive snow density and the total ther-mal resistance presented by the snow cover. This likely re-sults in a lesser effect on snow thermal conductivity than a

full treatment of the complex relationship between densityand conductivity, but our results still reflect the role of themore complex relationship. However, it is important to notethat because this method does not change the snow mass, thethermal diffusivity is not altered. As a result, the perturbationintroduced is a lesser effect than a fully bias corrected snowdepth would trigger.

Figure 4 presents the new conductivities used. While val-ues are calculated monthly, the values are interpolated at eachmodel time step. This avoids discontinuities in the conduc-tivity. Note that the higher conductivities in the CCSM biasexperiment compensate for the excess snow depth, and willallow for more thermal transfer through the snow–ice col-umn. In this way, the ice will experience the thermal effectof a thinner snow cover, correcting for the snow depth anddensity bias.

The sensitivity experiments are integrated over a period of60 yr. As the results indicate, this is sufficient to allow theCCSM and CICE to come to quasi-equilibrated states. Be-cause we use the SOM, there is no ocean to equilibrate, butby leaving CAM and CLM active the Arctic atmosphere andadjacent landmass are allowed to respond to changes in theice cover. In addition, we are able to assess the impacts of thechanges in sea ice on the overall Arctic climate, in particularthe overlying atmosphere.

4.2 CCSM bias sensitivity simulation results

We begin with a discussion of the changes to the ice stateduring the equilibrated period of the simulation, defined asthe last 20 yr of a 60 yr simulation. Following the discussionof the equilibrated state, this paper moves to the beginningof the simulation, where the transient changes in the ice statetriggered by our perturbations occur. We also examine theequilibrated changes in the Arctic atmosphere and generalclimate due to changes in the ice during the equilibrated pe-riod of the simulation, as well as the feedbacks between theatmosphere, ocean, and ice state.

An increase in thermal conductivity results in an increasein ice volume in the Arctic and corresponding increase in icearea (Fig. 5). This is expected because an increase in snowthermal conductivity allows increased energy flux throughthe snow–ice column, triggering increased basal ice forma-tion in the winter. Qualitatively, this result suggests that thechanges in winter ice growth exceed any changes in summerice melt caused by our modifications to snow thermal con-ductivity. These results confirm previous work by Fichefet etal. (2000). The addition of a fully coupled atmosphere in thisstudy allows us to extend the investigation of climate sen-sitivity, showing that the changes in ice conditions result inshifts in Arctic-wide temperature and cloud cover.

The equilibrated CCSM bias experiment produces 19 %more annual mean ice mass than the control and 7 % moreSeptember ice area than control, both significant (P < 0.05);see Fig. 5. This is a large effect, but is comparable to the 1m



Fig. 4. Intrannually varying now conductivity adjustments, appliedthroughout the CCSM bias experiment.

CICE-‐coupled Bias Vol m3 Coupled Control Vol m3

Bias Area km2 Control Area km2

Fig. 5. Arctic sea ice volume and extent for control run and CCSMbias experiment. Restricted to north of latitude 70◦ N.

increase seen in Holland et al. (2012). This suggests moreice formation occurring due to the increased transfer of heatthrough the snow–ice column, as would be expected giventhe results of Fichefet et al. (2000). However, as Fig. 6 shows,by the equilibrated period of model integration the experi-mental snow depth has increased significantly relative to thecontrol, which offsets the perturbation introduced and alsoinhibits conductive flux in the later period of model integra-tion, as elaborated in the following discussion. Therefore, itis likely that other mechanisms conducive to ice growth aretriggered by changes in the ice caused by the initial perturba-tion. We posit that this increase in snow depth is a result ofincreased ice survival, allowing more snow accumulation. Adiscussion of this effect is available in Blazey (2012).

Figure 7 documents the geographical patterns of changein ice thickness due to our snow thermal conductivity mod-ifications. Figure 7a shows the control simulation ice thick-ness, with the pattern of thickest ice near the Canadian ArcticArchipelago being typical (Bourke and Garrett, 1987). Fig-ure 7b shows the difference between control and the CCSMbias experiment. The thickest ice near the Canadian Arc-

Fig. 6. On-ice snow depth in the Arctic Ocean during equilibratedperiod of integration (last 20 yr of 60 yr simulation).

tic Archipelago experiences less pronounced changes in icegrowth, while the thinner ice along the marginal zone of theice pack experiences the greatest thickening. This conformsto expectations from Eq. (2). Due to the largerhice in the re-gion of thick ice, the thermal characteristics of the snow hasless impact on the total transfer of energy through the snow–ice column.

During the period of most rapid equilibration of the CCSMbias experiment (defined here as the first 20 yr of simula-tion following initialization) there is a statistically significant(P < 0.05) 7 % increase in annual mean conductive heat fluxat the top of the snow–ice surface in the study area, 70◦ Npoleward (see Fig. 8a). This 0.5 W m−2 increase in conduc-tive flux is consistent with the increase in snow thermalconductivity. This change is comparable to the 1.1 W m−2

change caused by the addition of black carbon and meltponds to CICE (Holland et al., 2011). In addition, note thatin the later portion of the integration this conductive fluxin the experimental simulation is lower than in the controlrun, likely due to increased ice and snow thickness as seenin Figs. 5, 6 and 7.

These shifts in conductive heat flux (Fig. 8a) have a signif-icant impact on surface conditions for the CCSM bias exper-iment, which results in an annual mean increase of 0.25 K insurface temperature over the control simulation (see Fig. 8b),which in turn contributes to an annual mean increase insurface long-wave flux from the surface of 0.8 W m−2 (seeFig. 8c). Both of these increases are significant (P < 0.05).

Figure 9 documents the annual mean difference in to-tal Arctic Ocean ice volume when comparing the experi-mental simulation to the control simulation. In Fig. 9, thesolid line (right axis) indicates the volume difference be-tween the two simulations is comparable until the modifiedthermodynamics begin to impact the mass balance aroundyear 25. However, the dotted lines in Fig. 9 (right axis) in-dicate that the ice mass balance terms are effected by thechanges in snow thermal conductivity immediately. In Fig. 9,



Fig. 7. Panel(a): equilibrated ice thickness (cm) and 15 % and 90 % ice extent in September (black contours) for control run. Panels(b):difference in ice thickness between CCSM bias experiment and the control run (cm) and 15 % and 90 % September ice extent (black contours).

A. Differenced conduc.ve flux (bias less control)

B. Differenced Surface temp. (bias less control)

C. Differenced long wave emiAed (bias less control)

Fig. 8. CCSM bias experiment surface energy budget terms relative to control in the study area (70◦ N poleward). Black lines indicate theannual mean difference in the energy budget term, righty axis. Year is from inception of the simulation, indicated on thex axis. Backgroundcolors indicate the monthly mean value for a given year, month indicated on the lefty axis. These monthly values are smoothed over afive-year window.(a) Surface conductive flux in W m−2 given by the top color bar; positive values indicate greater flux to the surface.(b)Surface temperature differences with respect to the control run K given by the top color bar.(c) Long-wave energy flux out in W m−2 givenby the bottom color bar.

the primary effect of the modified snow thermal conductiv-ity in the CCSM-coupled bias experiment is the increase incongelation ice, which is ice formed at the base of the seaice. Congelation ice forms when the conductive flux throughthe snow–ice column exceeds the flux from the ocean to theice during the winter months. This imbalance causes the re-lease of latent heat, and hence the formation of new ice atthe ocean-ice interface. This increased ice production is off-

set by a negative feedback mechanism through the increasedadvection of ice mass (because of thicker ice) from the ArcticBasin. However, near the end of the transient period of sim-ulation, around year 30, an increase in relative volume ten-dency associated with ocean melt (basal and lateral melt) oc-curs, indicative of the decreased transmission of energy fromthe ocean to either the base or margin of the ice pack. This isa result of the increased ice area seen in Fig. 5, which results



Fig. 9. Difference in ice volume and accumulated mass budgetterms during CCSM bias experiment. Year since inception of theexperiment is indicated on thex axis. Difference in annual aver-age ice volume (CCSM bias experiment less control) is representedby the solid line and lefty axis. Other lines and the rightx axisindicate cumulative difference in budget terms (CCSM bias exper-iment less control). A positive value indicates more ice is formedby a process or less melted in the bias experiment than in con-trol. Includes basal (congelation) ice growth, open water (frazil) icegrowth, basal/lateral (Ocean) melt, surface melt, and loss to trans-port out of the Arctic (advection).

in less exposed open water, so less solar shortwave radiationis absorbed by the ocean to be retransmitted to the ice: theice–ocean albedo feedback (Curry et al., 1995).

The flux differences during the last 20 yr of the 60 yr simu-lation are reported in Fig. 10; the initial perturbations to con-ductivity are fairly well equilibrated, with the differences be-tween experimental and control simulations becoming lessvariable, as with the total volume in Fig. 9. In addition, dueto reduced open water the flux in the summer ocean heat fluxdrops by as much as 20 W m−2. As a result, there is an an-nual average 6 W m−2 decrease in mean flux from the oceanto the ice (statistically significant (P < 0.05)). In addition tothe ice–ocean albedo feedback, an autumn decrease in long-wave flux from the atmosphere also occurs in the increasedconductivity simulation (see Fig. 10). This results in an an-nual mean of 2 W m−2 less long wave absorbed by the ice,also statistically significant (P < 0.05). Previous studies haveshown decreases in autumn ice area lead to an increase in au-tumn cloud cover, and represent an additional expected pos-itive ice area feedback (Schweiger et al., 2008).

Figure 11 examines the effects the perturbation has on theArctic atmosphere. While these effects are due to the changesin the snow cover, they are representative of other processes,such as the changed ocean heat flux initiated by the pertur-bation. Figure 11a shows a∼ 1 K drop in temperature nearthe surface in the CCSM bias sensitivity experiment. Thisdrop is significant for the lower atmosphere both in the an-nual mean and autumn. The initial perturbation reduces thethermal impedance through the snow–ice column, allowingfor more conductive flux and an initial increase in atmo-

LW up LW down Conduc,ve Flux at top of ice Ocean

CCSM bias experiment (70N): equilibrated flux difference

Fig. 10. Monthly mean difference in energy transfer between ex-perimental and control simulations during the equilibrated period(final 20 of 60 yr) for CCSM bias sensitivity. As in Fig. 10, positiveanomalies indicated more flux to the ice or less flux away from theice.

spheric surface temperature. If the ice were simply adjustingto the perturbation without feedbacks, the ice would thickenuntil the additional ice compensated for the thermally thin-ner snow cover. However, the decreased surface temperaturein the equilibrated CCSM bias experiment indicates feed-backs have resulted in an enhanced effect beyond, but dueto, the perturbation. In addition to the near-surface effect, theupper-atmosphere temperatures increase, but are not furtherdiagnosed in this study.

Figure 11b documents somewhat complex and generallynot significant (P < 0.05) reactions in cloud cover in our ex-perimental simulation. In the CCSM bias experiment the an-nual mean is only significant in the upper atmosphere, wherethere is a drop in cloud fraction. In the lower atmosphere,there is a significant drop in the autumn cloud fraction. FromFig. 11, the downwelling long-wave feedback observed inthe CCSM bias sensitivity experiment is the result of loweratmosphere decreases in both temperature and cloud cover.

In summary, an increase in snow thermal conductivity re-sults in significantly increased ice volume and area. In ad-dition, as seen in Fig. 7, ice of different thicknesses reactsdistinctly. We also stress the importance of oceanic and at-mospheric response and induced feedbacks to our perturba-tions. These findings indicate that the ice state is sensitiveto changes in snow conditions, in particular modifications tosnow thermal conductivity designed to be of the same mag-nitude as the depth biases identified in Sect. 3.

5 Conclusions

This study has evaluated the simulated on-ice snow depth anddensity in the Arctic Ocean, and investigated the impacts of



Fig. 11. Differenced (experiment less control) annual and autumn (September–November) mean temperature and cloud fractions for theequilibrated period of the CCSM bias sensitivity (last 20 of 60 yr) for 70◦ N poleward. Asterisks indicate the difference between experimentand control is significant (P < 05) at the height indicated.

snow depth bias on thermal conduction and in turn the icestate in coupled simulations. The on-ice snow depths pro-duced by CCSM are too thick. The model was found to besensitive to this bias, both in terms of ice characteristics andthe general Arctic climate.

The snow overlying the Arctic sea ice as simulated in aCCSM4 ensemble was found to be∼ 40 % too thick whencompared to in situ measurements. This bias increased to> 150 % during the summer months. In addition, the currentparameterization in CICE lacks a seasonal density evolution,resulting in excessive autumn and early winter snow densi-ties.

Following the identification of seasonally dependent snowdepth biases, the sensitivity experiment modified the ther-modynamic treatment of snow in CICE to compensate forthe biases. In this study, the thermal conductivity of thesnow was adjusted to simulate a thinner snowpack. The de-fault thermal conductivity of 0.3 W m−1 K−1 was increasedto ∼ 0.4 W m−1 K−1. This increase was seasonally variable,and accounted for the biases identified in both snow depthand density. The first-order examination of the sensitivity ofthe ice state to these biases revealed the anticipated response,whereby an increase in thermal conductivity, used to com-pensate for excessive snow depth in our CCSM bias sensi-tivity experiment, results in an increase in winter basal icegrowth (Fichefet et al., 2000). This increase results in bothincreased year-round ice volume and an increase in ice areaat the September sea ice area minimum.

However, this initial perturbation results in several feed-backs both within the ice and between the perturbed ice andthe other components of the Arctic climate. We determinedtwo limiting factors with regards to to continued increase in

ice volume. The increasing ice thickness serves to replace thethermal barrier lost due to the increased snow thermal con-ductivity. Increased thickness of ice advecting from the Arc-tic results in an enhanced sink on total ice volume. Positivefeedbacks resulting in an enhancement of the initial pertur-bation were also identified. Increased ice area resulted in acorresponding decrease in open water. In turn, this causes anincrease in average albedo, and a decrease in ice melt dueto melt by the ocean. Increased ice area results in decreasedlow-level clouds and reduced low-level temperatures, result-ing in reductions of the long-wave flux to the ice.

If the response to the bias adjustments were uniform,it would be tempting to regard the thermal conductivityas another tuning parameter. However, ice does not reactuniformly to changes in thermal conductivity. Specifically,thicker ice is less sensitive to changes in snow characteristics.This indicates that the response of the ice to snow character-istics is a complex system, and cannot simply be treated as aone-dimensional tuning parameter. In addition, the enhancedice area resulting from an increase in conductivity leads topositive feedbacks in the form of the ice–ocean feedback andice area – autumn cloud cover feedback.

These findings suggest that an improved understandingand treatment of the on-ice snow cover would enhance theability of CICE to generate an ice state consistent with ob-servations. In particular, the complex relation of ice thick-ness and snow thickness is important in light of observationsindicating changes in ice age and characteristics currently oc-curring in the Arctic. As sea ice thins, the snow is potentiallya larger component of the snow–ice column, increasing theimpact of biases in snow depth. Given the bias and model re-action described in this study, correction of the current bias



may result in slightly slower decline in sea ice than currentlyprojected. However, as the Arctic climate continues to shift,snow thickness is projected to decrease (Hezel et al., 2012),which leads to a complex and pronounced change in the ther-modynamics of the sea ice system (Blazey, 2012). Accord-ingly, accurate snow treatment in CICE is likely to play arole in accurate projects of the Arctic climate.

While the basic question of the importance of snow depthto sea ice has been posed in several previous studies, thosestudies are not in agreement. As such, this study is unique inasking whether a modern GCM is capable of producing theobserved snow depths, and what importance that capabilityhas, not only for the ice itself but also for the overall Arcticclimate. This study has determined that, likely due to biasesin precipitation and omission of snow processes, the GCMdoes not produce the correct on-ice snow depths. More im-portantly, it shows that not only the ice, but also the Arctic ingeneral, is affected by this bias. It shows that this is not a sim-ple linear correction, and the need for a careful assessment ofpotential improvements to the model to correct for this bias.This study demonstrates that, as models become more com-plex, there is need for targeted evaluation of the fidelity withwhich each model is able to reproduce the climate system.

Future work stemming from this study would likely focuson modifying the model treatment of snow to better repro-duce the snow depths observed in situ. The ideal in situ dataset for such an investigation would include coincident mea-surements of snow and ice thickness across the Arctic duringand after precipitation events, allowing for accurate param-eterization of the speed of snow redistribution. Tools suchas SnowModel (Liston and Elder, 2006) could also be usedto create a new set of parameters to simplify the redistribu-tion of snow over the sea ice and loss due to blowing snow.While the snow flux from the atmospheric model may con-tribute to the excessive snow identified in this study, Chunget al. (2010) found that the introduction of blowing snow intoa model of sea ice significantly reduced snow depth. The in-troduction of such a process would not only better representthe physical system, but also introduce a parameter that couldbe tuned to compensate for excess snow flux from the atmo-sphere.

Acknowledgements.The authors thank J. Maslanik for hissupport, both as faculty advisor and through funding supportthrough NSF Office of Polar Programs, grant OPP-0229473, andNASA Cryospheric Sciences Program, grants NNX07AR21G andNNX11AI48G. Funding from the Biological and EnvironmentalResearch division of the US Department of Energy Office of Sci-ence is gratefully acknowledged. Los Alamos National Laboratoryis operated by the National Nuclear Security Administration ofthe US Department of Energy under Contract No. DE-AC52-06NA25396.

Edited by: J. Stroeve

References

Arctic Climatology Project: Environmental working group Arcticmeteorology and climate atlas, edited by: F. Fetterer and V. Ra-dionov, Boulder, CO, National Snow and Ice Data Center, CD-ROM, 2000.

Bailey, D. C., Hannay, M„ Holland, R., and Neale, N. C. A. R.:Slab Ocean Model Forcing, NCAR,http://www.cesm.ucar.edu/models/ccsm4.0/data8/SOM.pdf\mskip\medmuskip, (lastaccess: 10 April 2012) 2012.

Bitz, C. M., Shell, K. M., Gent, P. R., Bailey, D., Danabasoglu, G.,Armour, K. C., Holland, M. M., and Kiehl, J. T.: Climate Sen-sitivity in the Community Climate System Model Version 4, J.Climate, 25, 3053–3070, doi:10.1175/JCLI-D-11-00290.1, 2011.

Bitz, C. M. and Lipscomb, W. H.: An energy-conserving thermody-namic model of sea ice, J. Geophys. Res., 104, 669–677, 1999.

Blazey, B. A.: Snow Cover on the Arctic Sea Ice: Model Valida-tion, Sensitivity, and 21st Century Projections, Ph.D. Disserta-tion, University of Colorado, USA, 2012.

Bourke, R. H. and Garret, R. P.: Sea ice thickness distribution in theArctic Ocean, Cold Regions Sci. Technol., 13, 259–280, 1987.

Briegleb, B. P., Danabasoglu, G., and Large, W. G.: An Over-flow Parameterization for the Ocean Component of the Com-munity Climate System Model NCAR Tech, Note NCAR/TN-481+STR, Draft, 72 pp., 2010.

Briegleb, B. P. and Light B.: A Delta-Eddington Multiple Scatter-ing Parameterization for Solar Radiation in the Sea Ice Compo-nent of the Community Climate System Model, Ncar Tech. NoteNCAR/TN-472+STR, 100 pp., 2007.

Brown, R. D. and Cote, P.: Interannual variability of landfast icethickness in the Canadian High Arctic, 1950-8, Arctic, 45, 273–284, 1992.Cheng, B., Zhang, Z., Vihma, T., Johansson, M., Bian, L., Li, Z.,and Wu, H.: Model experiments on snow and ice thermodynam-ics in the Arctic Ocean with CHINARE 2003 data, J. Geophys.Res., 113, C09020, doi:10.1029/2007JC004654, 2008.

Chung, Y.-C., Bélair, S., and Mailhot, J.: Blowing Snow on ArcticSea Ice: Results from an Improved Sea Ice-Snow-Blowing SnowCoupled System, J. Hydrometeor., 12, 678–689, 2011.

Curry, J. A., Schramm, J., and Ebert, E. E.: On the sea ice albedoclimate feedback mechanism, J. Climate, 8, 240–247, 1995.

Danabasoglu, G., Large, W. G., and Briegleb, B. P.: Climate impactsof parameterized Nordic Sea overflows, J. Geophys. Res., 115,C11005, doi:10.1029/2010JC006243, 2010.

De Boer, G., Chapman, W., Kay, J. E., Medeiros, B., Shupe, M.D., Vavrus, S., and Walsh, J.: A Characterization of the Present-Day Arctic Atmosphere in CCSM4, J. Climate, 25, 2676–2695,doi:10.1175/JCLI-D-11-00228.1, 2012.

Déry, S. J. and Tremblay, L.-B. Modeling the effects of wind re-distribution on the snow mass budget of polar sea ice, J. Phys.Ocean., 34, 258–271, 2004.

Fichefet, T., Tartinville, B., and Goosse, H.: Sensitivity of theAntarctic sea ice to the thermal conductivity of snow, Geophys.Res. Lett., 27, 401–404, 2000.

Fichefet, T. and Morales Maqueda, M. A.: Modeling the influenceof snow accumulation and snow-ice formation on the seasonalcycle of the Antarctic sea-ice cover, Clim. Dyn., 15, 251–268,1999.

Flanner, M. G., Zender, C. S., Randerson, J. T., and Rasch,P. J.: Present-day climate forcing and response from


http://www.cesm.ucar.edu/models/ccsm4.0/data8/SOM.pdfmskip medmuskip

http://www.cesm.ucar.edu/models/ccsm4.0/data8/SOM.pdfmskip medmuskip

http://dx.doi.org/10.1175/JCLI-D-11-00290.1

http://dx.doi.org/10.1029/2007JC004654

http://dx.doi.org/10.1029/2010JC006243



black carbon in snow, J. Geophys. Res., 112, D11202,doi:10.1029/2006JD008003, 2007.

Gent, P. R., Danabasoglu, G., Donner, L. J., Holland, M. M., Hunke,E. C., Jayne, S. R., Lawrence, D. M., Neale, R. B., Rasch, P.J., Vertenstein, M., Worley, P. H., Yang, Z.-L., and Zhang, M.:The Community Climate System Model version 4, J. Climate,24, 4973–4991, doi:10.1175/2011JCLI4083.1, 2011.

Hezel, P. J., Zhang, X., Bitz, C. M., Kelly, B. P., and Massonnet, F.:Projected decline in spring snow depth on Arctic sea ice causedby progressively later autumn open ocean freeze-up this century,Geophys. Res. Lett., 39, L17505, doi:10.1029/2012GL052794,2012.

Holland, D. M, Mysak, L. A., Manak, D. K., Oberhuber, J. M.: Sen-sitivity Study of a Dynamic Thermodynamic Sea Ice Model, J.Geophys. Res. 98, 2561–2586, 1993.

Holland, M. M., Bailey, D. A., Briegleb, B. P.,Light, B., and Hunke,E.: Improved Sea Ice Shortwave Radiation Physics in CCSM4:The Impact of Melt Ponds and Aerosols on Arctic Sea Ice, J.Climate, 25, 1413–1430, doi:10.1175/JCLI-D-11-00078.1, 2012.

Holland, M. M., Bitz, C. M., Hunke, E. C., Lipscomb, W. H., andSchramm, J. L.: Influence of the sea ice thickness distribution onpolar climate in CCSM3, J. Climate, 19, 2398–2414, 2006a.

Holland, M. M., Bitz, C. M., and Tremblay, B.: Future abrupt re-ductions in the summer Arctic sea ice. Geophys. Res. Lett., 33,L23503, doi:10.1029/2006GL028024, 2006b.

Hunke, E. C. and Lipscomb, W. H.: CICE: the Los Alamos Sea IceModel, Documentation and Software User’s Manual Version 4.1.T-3 Fluid Dynamics Group, Los Alamos National Laboratory,Tech. Rep. LA-CC-06-012, 2010.

Hunke, E. C. and Dukowicz J. K.: The Elastic-Viscous-Plastic SeaIce Dynamics Model in General Orthogonal Curvilinear Coordi-nates on a Sphere – Incorporation of Metric Terms. Mon. Wea.Rev., 130, 1848–1865, 2002.

Jahn, A., Bailey, D. A., Bitz, C. M., Holland, M. M., Hunke, E. C.,Kay, J. E., Lipscomb, W. H., Maslanik, J. A., Pollak, D., Sterling,K., Strœve, J.: Late 20th Century Simulation of Arctic Sea Iceand Ocean Properties in the CCSM4, J. Climate, 25, 1431–1452,doi:10.1175/JCLI-D-11-00201.1, 2012.

Lawrence, D. M., Oleson, K. W., Flanner, M. G., Thornton, P. E.,Swenson, S. C., Lawrence, P. J., Zeng, X., Yang, Z.-L., Levis, S.,Sakaguchi, K., Bonan, G. B., and Slater, A. G.: Parameterizationimprovements and functional and structural advances in version4 of the Community Land Model, J. Adv. Model. Earth Sys., 3,27 pp., doi:10.1029/2011MS000045, 2011.

Lecomte, O., Fichefet, T., Vancoppenolle, M., and Nicolaus, M.: Anew snow thermodynamic scheme for large-scale sea-ice models,Ann. Glaciol., 52, 337–346, 2011.

Lin, S. J. : A “Vertically Lagrangian” finite-volume dynamical corefor global models, Mon. Wea. Rev., 132, 2293–2307, 2004.

Lipscomb, W. H. and Hunke, E. C.: Modeling sea ice transport us-ing incremental remapping, Mon. Wea. Rev., 132, 1341–1354,2004.

Lipscomb, W. H.: Remapping the thickness distribution in seaice models, J. Geophys. Res.-Oceans, 106, 13989–14000,doi:10.1029/2000JC000518, 2001.

Liston, G. E. and Elder, K.: A distributed snow-evolution modelingsystem (SnowModel), J. Hydrometeorol., 7, 1259–1276, 2006.

Maslanik, J., Stroeve, J., Fowler, C., and Emery, W.: Distributionand trends in Arctic sea ice age through spring 2011, Geophys.Res. Lett., 38, L13502, doi:10.1029/2011GL047735, 2011.

Maslanik, J. A., Fowler, C., Stroeve, J., Drobit, S., Zwally, J., Yi, D.,and Emery, W.: A younger, thinner Arctic ice cover: Increasedpotential for rapid, extensive sea-ice loss, Geophys. Res. Lett.,34, L24501, doi:10.1029/2007GL032043, 2007.

Massom, R. A., Eicken, H., Haas, C., Jeffries, M. O., Drinkwater,M. R., Sturm, M., Worby, A. P., Wu, X., Lytle, V. I., Ushio, S.,Morris, K., Reid, P. A., Warren, S., and Allison, I.: Snow onAntarctic sea ice, Rev. Geophys., 39, 413–445, 2001.

Maykut, G. A. and Untersteiner N.: Some results from a time de-pendent, thermodynamic model of sea ice, J. Geophys. Res., 76,1550–1575, 1971.

Meier, W. N., Stroeve, J., and Fetterer, F.: Whither Arctic sea ice?A clear signal of decline regionally, seasonally, and extendingbeyond the satellite record, Ann. Glaciol., 46, 428–434, 2007.

Neale, R. B., Richter, J., Park, S., Lauritzen, P. H., Vavrus, S. J.,Rasch, P. J., and Zhang, M.: The Mean Climate of the Commu-nity Atmosphere Model (CAM4) in Forced SST and Fully Cou-pled Experiments, J. Climate, 26, 5150–5168, doi:10.1175/JCLI-D-12-00236.1, 2013.

NSIDC: A better year for the cryosphere,http://nsidc.org/arcticseaicenews/2013/10/a-better-year-for-the-cryosphere/,(last access: 3 Oct 2013), 2013.

NSIDC Daily Image: Update 2012, National Snow and Ice DataCenter, ,http://nsidc.org/arcticseaicenews/2012/05/daily-image/\mskip\medmuskip, (last access: 18 Sept 2012), 2012.

Oleson, K. W., Lawrence, D. M., Bonan, G. B., Flanner, M. G.,Kluzek, E., Lawrence, P. J., Levis, S., Swenson, S. C., Thornton,P. E., Dai, A., Decker, M., Dickinson, R., Feddema, J., Heald,C. L., Hoffman, F., Lamarque, J. F., Mahowald, N., Niu, G.-Y.,Qian, T., Randerson, J., Running, S., Sakaguchi, K., Slater, A.,Stockli, R., Wang, A., Yang, Z.-L., Zeng, X., and Zeng, X.: Tech-nical description of version 4.0 of the Community Land Model,NCAR Tech. Note NCAR/TN-478+STR, 257, 2010.

Parkinson, C. L., Cavalieri, D. J., Gloersen, P., Zwally, H. J., andComiso, J. C.: Arctic sea ice extents, areas, and trends, 1978–1996, J. Geophys. Res., 104, 20837–20856, 1999.

Perovich, D. K., Richter-Menge, J. A., Elder, B., Claffey, K., andPolashenski, C.: Observing and understanding climate change:Monitoring the mass balance, motion, and thickness of Arcticsea ice,http://IMB.crrel.usace.army.mil(last access:04.12.2013),2009.

Perovich, D. K., Nghiem, S. V., Markus, T., and Schweiger, A.: Sea-sonal evolution and internannual variability of the local energyabsorbed by the Arctic sea ice-ocean system, J. Geophys. Res.,112, D11202, doi:10.1029/2006JC003558, 2007.

Perovich, D. K., Grenfell, T. C., Light, B., and Hobbs, P. V.: Sea-sonal evolution of the albedo of multiyear Arctic sea ice, J. Geo-phys. Res., 107, 8044, doi:10.1029/2000JC000438, 2002.

Perovich, D. K., Roesler, C. S., and Pegau, W. S.: Variability in seaice optical properties, J. Geophys. Res., 103, 1193–1209, 1998.

Schweiger, A. J., Lindsay, R. W., Vavrus, S., and Francis, J. A.: Re-lationships between Arctic Sea Ice and Clouds during Autumn,J. Climate, 21, 4799–4810, doi:10.1175/2008JCLI2156.1, 2008.

Smith, R., Jones, P., Briegleb, B., Bryan, F., Danabasoglu, G., Den-nis, J., Dukowicz, J., Eden, C., Fox-Kemper, B., Gent, P., Hecht,M., Jayne, S., Jochum, M., Large, W., Lindsay, K., Maltrud, M.,


http://dx.doi.org/10.1175/2011JCLI4083.1

http://dx.doi.org/10.1029/2012GL052794


http://dx.doi.org/10.1029/2006GL028024


http://dx.doi.org/10.1029/2011MS000045

http://dx.doi.org/10.1029/2011GL047735

http://dx.doi.org/10.1029/2007GL032043



http://nsidc.org/arcticseaicenews/2013/10/a-better-year-for-the-cryosphere/

http://nsidc.org/arcticseaicenews/2013/10/a-better-year-for-the-cryosphere/

http://nsidc.org/arcticseaicenews/2012/05/daily-image/mskip medmuskip

http://nsidc.org/arcticseaicenews/2012/05/daily-image/mskip medmuskip

http://IMB.crrel.usace.army.mil

http://dx.doi.org/10.1029/2006JC003558

http://dx.doi.org/10.1029/2000JC000438

http://dx.doi.org/10.1175/2008JCLI2156.1


Norton, N., Peacock, S., Vertenstein, M., and Yeager, S.: The Par-allel Ocean Program (POP) Reference Manual: Ocean Compo-nent of the Community Climate System Model (CCSM), LAUR-10-01853, 2010.

Stroeve, J. C., Kattsov, V., Barrett, A., Serreze, M., Pavlova, T.,Holland, M., and Meier, W. N.: Trends in Arctic sea ice extentfrom CMIP5, CMIP3 and observations, Geophys. Res. Lett., 39,L16502, doi:10.1029/2012GL052676, 2012.

Stroeve, J., Holland, M. M., Meier, W., Scambos, T., and Serreze,M.: Arctic sea ice decline: Faster than forecast, Geophys. Res.Lett., 34, L09501, doi:10.1029/2007GL029703, 2007.

Sturm, M. and Liston, G.E.: The Snow Cover on Lakes of the ArcticCoastal Plain of Alaska, J. Glaciol., 49, 370–380, 2003.

Sturm, M., Perovich, D. K., and Holmgren, J.: Thermal conductivityand heat transfer through the snow and ice of the Beaufort Sea. J.Geophys. Res.-Oceans, 107, 8043, doi:10.1029/2000JC000409,2002.

Sturm, M., Holmgren, J., König, M., and Morris, K.: The thermalconductivity of seasonal snow, J. Glaciol., 43, 26–41, 1997.

Thorndike, A. S., Rothrock, D. A., Maykut, G. A., and Colony, R.:The thickness distribution of sea ice, J. Geophys. Res., 80, 4501–4513, 1975.

Vavrus, S. J., Holland, M. M., Jahn, A., Bailey, D. A., and Blazey, B.A.: 21st-Century Arctic Climate Change in CCSM4, J. Climate,25, 2696–2710, doi:10.1175/JCLI-D-11-00220.1, 2012.

Vavrus, S. and Waliser, D.: An improved parameterization for simu-lating Arctic cloud amount in the CCSM3 climate model, J. Cli-mate, 21, 5673–5687, 2008.

Warren, S. G., Rigor, I. G., Radinov, V. F., Bryazgin, N. N., Alek-sandrov, Y. I., and Colony, R.: Snow depth on Arctic ice, J. Clim.,9, 1814–1829, 1999.

Wu, X., Budd, W. F., Lytle, V. I., and Massom, R. A.: The effect ofsnow on Antarctic sea ice simulations in a coupled atmosphere-sea ice model, Clim. Dyn, 15, 127–143, 1999.


http://dx.doi.org/10.1029/2012GL052676

http://dx.doi.org/10.1029/2007GL029703

http://dx.doi.org/10.1029/2000JC000409


arctic ocean sea ice snow depth evaluation and bias ... · abstract. sea ice cover in the arctic...

Documents